Kenneth Glad
Technology constantly promises to make hard work disappear with the click of a button. We want machines to do the heavy lifting and hand us the final answers right away. But sometimes the physical world refuses to cooperate.
Researchers recently turned to artificial intelligence to solve a massive historical puzzle. A team deployed three different deep learning models across the vast Nefud desert in Saudi Arabia. Their goal was simple. They wanted to see if software could spot ancient stone structures hidden in the sand.
Looking for these ruins usually means spending years walking the landscape or staring at satellite photos. The new goal was speed.
“Is there a faster way to do it? Can we automate this process and get software to tell us where they all are?” Maria Guagnin told HistorieNet. As a researcher at the Max Planck Institute of Geoanthropology, she wondered if computers could take over the tedious visual work.
“Or should we have trained archaeologists look at the pictures and find them?” she added.
Finding the familiar
The results were a mixed bag. When it came to large, rectangular Stone Age monuments, the computers succeeded easily. But smaller stone mounds and complex tombs completely baffled the machines.
This exposes a glaring limitation in the technology. While algorithms excel at recognizing existing patterns, they cannot replicate human curiosity.
“Right now we can only train a software to identify things we already know. Therefore they only learn known structure types and find more of the same,” Guagnin said in the interview.
She noted that real scientific breakthroughs come from the unexpected. “But the interesting thing in research is the structure types that are not yet known. You cannot train the systems on things that are not yet known,” Guagnin said.
Need for human eyes
Other experts agree that machines will not conquer the field anytime soon. At Johannes Gutenberg University Mainz, archaeologist Maxime Brami relies on laser scanning to find ancient Roman sites hidden in European forests.
He explained to HistorieNet that computers get confused easily:
“For example, haystacks can look like burial mounds in LiDAR images. Therefore, archaeologists still have to visit the potential site in person.”
While AI speeds up the early search phases, flesh-and-blood humans must verify the findings.
“In this respect, AI is far from being able to replace archaeologists. It simply acts as a tool to assist them in their research and help them identify potential sites more efficiently,” Brami concluded.
Source: Historienet
